Artificial intelligence (AI) holds a lot of potential for all types of industries and across different sizes of organizations.
A new report released today by global IT services vendor Infosys puts a dollar figure on that potential: According to the Infosys Data+AI Making AI Real report, a total of $467 billion in incremental profit can be made by global organizations if they are able to optimize AI and data practices.
The forecast from the Infosys Knowledge Institute, which is the research division of Infosys, was based in part on a survey of 2,500 IT leaders in 13 industries across the U.S., U.K., France, Germany, Australia and New Zealand. The key challenge that the report identified is not that organizations need to start using AI, but rather they need to integrate best practices for AI — as well as data — to actually get the full benefits of the technology.
“We were surprised to find that 80% of companies put their first AI system in production in the past four years,” Mohit Joshi, president of Infosys, told VentureBeat. “AI is not new, but companies are new to deploying their own AI systems and this contributes to the basic capabilities and low levels of satisfaction we found.”
There is no shortage of analysis and industry reports that discuss the potential for AI.
A March 2022 report from Cloudera found that 77% of surveyed knowledge workers believed AI, machine learning (ML) and data analytics would benefit their organizations over the next three years. Deloitte issued a report in October in which 94% of business leaders agreed that AI would be critical to success in the coming years. The Deloitte report also identified underachievers, with 22% of organizations deploying AI and getting low outcomes.
The Infosys report had a much starker figure on underachievement with AI. A scant 26% of organizations across all surveyed industries claimed they were “highly satisfied with their AI and data tools.” There are a number or reasons for the low level of satisfaction, including the need for more automation to help detect potential issues, as well as regulatory compliance. The report also found that the majority of AI models (63%) are relatively basic and require humans to make them work properly. The basic models are also plagued by a lack of good data practices, including data verification. There is also a clear need for more responsible AI and bias mitigation capabilities.
“Building robust AI ethics and bias management practices significantly increases satisfaction with AI,” Joshi said. “Companies must also think differently about data – the most instructive data for a particular AI case could come from a third party or publicly available data.”
Satisfaction with AI is not uniform across all industries — some are doing better than others. Financial service businesses are the most satisfied with their AI as an industry, Joshi said, while in terms of specific use cases, manufacturers and high-tech companies showed the highest satisfaction with using AI for predictive maintenance and reducing downtime in systems.
In order for organizations to fully realize the potential of AI and pull in the incremental revenues that Infosys estimates, there is a need to enable data sharing and data trust, as well as making AI a business priority and not just a task for data scientists.
The Infosys report identified that organizations that have mature data management and collaboration practices have better business outcomes from their data. Among the organizations that are highly satisfied with their AI tools are those that have responsible data practices that help to ensure that data is trustworthy.
Extending AI teams beyond data scientists is critical to success and helps enable alignment with business objectives. Joshi said it’s important that AI teams also include end users and experts in the business problems that need to be solved.
“Our analysis shows it’s most critical to always include the business expert and data scientist on the AI team,” Joshi said.
When it comes to moving to more advanced AI capabilities, Infosys has developed a taxonomy known as Sense, Understand, Respond, Evolve (SURE). The Sense capability is about being able to identify patterns, while Understand is about being able to make predictions. Respond is defined as the ability to also act autonomously, while the Evolve step is about models that train and improve themselves. Joshi said that he’s hopeful that over time, more organizations will uplevel from basic Sense and Understand capabilities to the more advanced and autonomous Respond and Evolve tiers.
“If companies can build greater trust in AI, we will see improving rates of satisfaction with data and AI,” Joshi said. “This is critical because companies with higher satisfaction have more effective AI systems.”
Copyright: VentureBeat Integrated AI can generate $460 billion in incremental profits, report finds | VentureBeat